Polyphosphate Functions In Vivo as an Iron Chelator and Fenton Reaction Inhibitor

How do organisms deal with free iron? On the one hand, iron is an essential metal that plays crucial structural and functional roles in many organisms. On the other hand, free iron is extremely toxic, particularly under aerobic conditions, where iron rapidly undergoes the Fenton reaction and produces highly reactive hydroxyl radicals. Our study now demonstrates that we have discovered one of the first physiologically relevant nonproteinaceous iron chelators and Fenton inhibitors. We found that polyphosphate, a highly conserved and ubiquitous inorganic polyanion, chelates iron and, through its multivalency, prevents the interaction of iron with peroxide and therefore the formation of hydroxyl radicals. We show that polyP provides a crucial iron reservoir for metalloproteins under nonstress conditions and effectively chelates free iron during iron stress. Importantly, polyP is present in all cells and organisms and hence is likely to take on this crucial function in both prokaryotic and eukaryotic cells.

lines revealed that tumorous cells gain resistance against cisplatin by increasing their levels of cysteine-enriched peptides (i.e., glutathione) and proteins (i.e., metallothioneins), which capture cisplatin before it reaches the DNA (12)(13)(14). This cellular response also seems to aid in mitigating oxidative stress, a frequently observed side effect of cisplatin treatment (15).
Recent studies from our lab revealed that upon cisplatin treatment, cancer cells drastically upregulate and redistribute their levels of inorganic polyphosphate (polyP) (16). PolyP is a polymer of up to 1,000 inorganic phosphate (P i ) molecules, linked by high-energy phosphoanhydride bonds (reviewed in references 17). In bacteria, polyP protects against a variety of different stress conditions (i.e., oxidative stress, heat stress), stimulates biofilm formation, and regulates virulence (18,19). Some of these functions can be explained by the ability of polyP to work as a protein-stabilizing scaffold (19). As such, polyP protects soluble proteins against stress-induced aggregation while promoting the formation of functional amyloids, including those involved in biofilms (20). Other potential functions that have been associated with polyP are based on its chemical features as a buffer or high-energy storage molecule (reviewed in reference 21).
To gain more insights into the working mechanism of cisplatin and the role that polyP might play in the cellular response to this drug, we compared the effects of cisplatin treatment on wild-type and polyP-deficient Escherichia coli. Our studies demonstrated that cisplatin triggers a gene expression pattern in wild-type bacteria that is consistent with the inactivation of the repressor Fur, the master regulator of iron homeostasis (22). The resulting gene expression changes lead to an apparent increase in iron uptake, whose deleterious effect is effectively mitigated by endogenous polyP. Deletion of the polyP-synthesizing machinery causes a dramatic increase in cisplatininduced mutagenesis rate and cell death. Both phenotypes are fully prevented in polyP-depleted bacteria by overexpressing iron storage proteins or by globally reducing the number of iron-containing proteins. Subsequent in vivo and in vitro studies revealed that polyP acts as a hitherto unknown iron-storage molecule under both stress and nonstress conditions and, by chelating labile iron, acts as a physiologically relevant Fenton inhibitor.

RESULTS
PolyP protects E. coli against cisplatin toxicity. Despite its prevalent use in antitumor treatment and its known broad antibacterial activity, the exact mechanism by which cisplatin kills cells is still not fully understood. To obtain more detailed insight into the cellular effects of cisplatin, we investigated the responses to and defenses against cisplatin toxicity in bacteria. Based on our recent discovery that polyP serves as an active defense mechanism against oxidative protein damage in bacteria (19), we compared the cisplatin sensitivity of wild-type E. coli with mutant strains that lack either the polyP-synthesizing polyP kinase (PPK) or the polyP-hydrolyzing enzyme exopolyphosphatase (PPX). Whereas the ppk deletion strain is unable to make polyP, the ppx deletion strain accumulates slightly higher polyP levels than wild-type E. coli under normal growth conditions, and significantly higher polyP levels than wild type under both nutrient shift and oxidative stress conditions (19,23). We grew all three strains to mid-log phase in minimal morpholinepropanesulfonic acid (MOPS) glucose (MOPS-G) medium and exposed them to increasing amounts of cisplatin either on plates (Fig. 1A) or in liquid culture (see Fig. S1A in the supplemental material). Whereas E. coli wild type or the ppx deletion mutant strain did not show any growth defects when incubated on plates supplemented with 4 g/ml of cisplatin, the ppk deletion mutant, which lacks detectable polyP levels (19), showed a reproducible 3-to 4-log reduction in cell survival (Fig. 1A). We obtained very similar results when we conducted the experiments in liquid culture. After a 20-h treatment with 10 g/ml cisplatin in liquid medium, wild-type E. coli showed an about 3-log decrease in survival while the ppk deletion strain showed a greater than 6-log decrease (Fig. S1A).
Analysis of the endogenous polyP levels in wild-type E. coli did not reveal any significant upregulation of polyP in response to cisplatin treatment (Fig. S1B), suggesting that the steady-state levels of polyP that are present in stationary-phase E. coli (24) are sufficient to confer the observed protection. To ascertain, however, that polyP synthesis is indeed required for the observed cisplatin resistance in wild-type E. coli, we transformed the ppk deletion strain with plasmids encoding either the native PPK protein or the previously characterized, catalytically inactive variant PPK-H435A (25). As shown in Fig. 1B, the expression of wild-type PPK fully rescued the cisplatin sensitivity of the ppk deletion strain. The expression of the catalytically inactive PPK variant, on the other hand, failed to rescue the growth defect. These results strongly suggested that endogenous levels of polyP are necessary and sufficient to protect E. coli against chronic cisplatin stress.
PolyP protects E. coli against cisplatin-induced DNA damage. To begin to understand how polyP protects bacteria against cisplatin toxicity, we analyzed the expression of select heat shock and SOS response genes in cisplatin-treated wild-type and ppk and ppx deletion mutant strains. This line of experiments was instigated by our previous study, which showed that during severe oxidative stress, polyP-deficient bacteria upregulate their heat shock gene expression levels in an apparent attempt to compensate for the lack of polyP's chaperone function (19). Analysis of the mRNA levels of IbpA and DnaK, two genes whose expression levels are highly responsive toward protein unfolding stress in E. coli (26), did not reveal any significant change upon cisplatin treatment in any of the three tested strains (Fig. S1C). In contrast, however, we found that the mRNA levels of the gene encoding the cell division inhibitor SulA, a major component of the SOS response and an inducer of bacterial filamentation, significantly increased upon cisplatin treatment in the wild-type E. coli and ppx deletion strain and went up even more in bacteria lacking polyP. These results agreed with the original observation that cisplatin treatment triggers bacterial filamentation (27) and indicated that at the concentrations used, cisplatin works as a DNA rather than a protein-damaging reagent in bacteria.
To assess the levels of DNA damage that cisplatin elicits in the ppk deletion strain versus in bacteria that contain measurable levels of polyP, we determined the mutagenesis rates before and after cisplatin treatment by counting the number of bacteria that gain the ability to grow on rifampin plates (28). Rifampin-resistant mutations in the Logarithmically growing E. coli wild type containing the empty vector pBAD18b, or the ppk mutant strain containing either the empty vector or expressing wild-type PPK or the enzymatically inactive PPK H435A mutant, was 10-fold serially diluted, spotted onto plates containing 0.02% (wt/vol) arabinose and 4 g/ml cisplatin, and incubated at 37°C overnight. Experiments shown in panels A and B were conducted at least 4 times, and a representative result is shown. (C) The mutagenesis rate of a logarithmically growing wild-type (wt), ⌬ppk, or ⌬ppx strain or a ⌬ppk strain expressing the highly active PPK-D230N mutant protein before and 1 h after treatment with 10 g/ml cisplatin in liquid culture was determined by counting the number of colonies able to grow on rifampin plates (n ϭ 3; *, P Ͻ 0.05; ****, P Ͻ 0.0001; ns, nonsignificant, one-way ANOVA).
RNA polymerase gene rpoB arise from single base substitutions, which prevent rifampin from binding to and hence inhibiting RNA polymerase (29). Under nonstress conditions, the spontaneous rate of mutagenesis in all three tested strains was similarly low (Ͻ10 mutants per 10 8 cells) (Fig. 1C). Not unexpectedly given cisplatin's mode of action, this number drastically increased upon cisplatin treatment. However, the mutagenesis rate of the ppk deletion was almost 3-fold higher than the mutagenesis rates of cisplatintreated wild-type E. coli or the ppx deletion strain. Expression of a highly active PPK variant, PPK-D230N, which substantially increases the steady-state levels of polyP in vivo (23), reduced the mutagenesis rate of the ppk deletion strain to levels that were even lower than the ones observed in cisplatin-treated wild-type E. coli (Fig. 1C). These results suggested a previously unrecognized function of polyP in bacterial DNA damage control.
Cisplatin triggers extensive transcriptional changes in wild-type and polyPdepleted E. coli. We conducted transcriptome sequencing (RNA-seq) analysis to compare the transcriptional profile of wild-type and ⌬ppk cells in response to a sublethal dose of cisplatin (i.e., 20 g/ml cisplatin for 15 min). Compared to the respective untreated controls, we identified more than 1,100 differentially expressed genes (DEGs) in each strain (see Tables S1 to S4). We categorized the DEGs into clusters of ontology according to their functional annotation (GOterm) (30) and focused our primary analysis on the two following groups: (i) DEGs in wild-type E. coli upon cisplatin treatment ( Fig. 2A; Tables S1 and S4), since we reasoned that those genes will likely reveal the cellular effects that cisplatin exerts in bacteria; and (ii) DEGs in cisplatin-treated wild type versus Δppk strain ( Fig. 2B; Tables S3 and S4), since we deduced that those genes will likely provide mechanistic insights into how polyP protects bacteria against cisplatin stress. The most upregulated genes in cisplatin-treated wild-type E. coli compared to the untreated control belonged to members of the SOS response ( Fig. 2A and Table S1). This was not an unexpected result since the SOS response is the canonical response to DNA damage. It controls more than 40 genes involved in transcriptional regulation, DNA repair mechanism, cell cycle arrest, and error-prone DNA synthesis (31-33) ( Fig. 2A). Many of the next most significantly altered cluster of DEGs in cisplatin-treated wild-type E. coli included genes involved in iron-sulfur cluster assembly, iron uptake, enterobactin synthesis, and iron regulation ( Fig. 2A, red bars). These results suggested that cisplatin treatment causes a disturbance in the cellular iron homeostasis. Indeed, when we compared the DEGs in cisplatin-treated wild type with the DEGs identified under iron starvation or repletion conditions (22), we observed a statistically highly significant overlap (Fig. 2C). Particularly striking was the overlap between genes differentially regulated by cisplatin and genes previously reported to be directly controlled by the master transcriptional repressor ferric uptake regulator (FUR) (22) (Fig. 2C). In fact, the expression pattern in cisplatin-treated wild-type bacteria strongly resembled the expression pattern in E. coli mutants lacking functional FUR, i.e., the upregulation of genes involved in iron uptake and the selective downregulation of genes involved in iron utilization (Fig. 2D). These results suggested that cisplatin treatment affects iron homeostasis in wild-type bacteria, either by triggering or by signaling an iron starvation response. Other significantly enriched gene clusters in cisplatin-treated wild-type bacteria included genes involved in fermentation and anaerobic electron transport chain (ETC), as well as nitrate assimilation and sulfur and phosphate metabolism ( Fig. 2A).
Cisplatin elicits iron stress in polyP-deficient E. coli. Direct comparison of the DEGs in cisplatin-treated wild-type E. coli and polyP-deficient mutant cells revealed numerous gene clusters that responded similarly to cisplatin treatment in the two strains (compare Fig. 2A with Fig. S2A; Table S3). In stark contrast, however, we found dramatic differences between the two strains in the expression of genes involved in amino acid transport, translation, transposition, and, particularly, iron homeostasis (Fig. 2B). Given the close connection between iron, oxidative stress, and DNA damage, we subsequently focused on the DEGs associated with iron homeostasis. We observed that polyP-depleted cells respond to cisplatin treatment with a much less pronounced expression of iron uptake genes and an even more pronounced downregulation of iron utilization genes ( Fig. 2D and Table S3). This expression pattern suggested that polyPdepleted bacteria experience a relative increase in the intracellular labile iron pool upon    (Tables S3 and S4). P values are shown in white on each bar, from a modified Fisher exact test. (C) Venn diagram of genes belonging to the fur regulon (green circle), genes differentially expressed in wild-type E. coli upon cisplatin treatment (cyan circle), and genes differentially expressed in a fur deletion strain under iron repletion and starvation conditions (22) (red circle). P values and odds ratios from Fisher exact tests. (D) Heatmap of Fur-regulated genes, which are differentially expressed in cisplatin-treated wild-type (wt) and ⌬ppk cells. Data are log 2 fold change, range Ϫ4 (blue) to 4 (red). Ratio of ppk to wild-type cells is shown in the right column. Genes are organized according to their functional annotations. cisplatin treatment, which would explain their relative decrease in survival and increase in mutagenesis rate compared to wild-type bacteria ( Fig. 1A and C). Given the polyanionic structure of polyP, and its previously shown ability to interact with divalent metals, such as Ca 2ϩ , Mg 2ϩ , and certain heavy metals (34,35), we therefore considered the possibility that polyP serves as a hitherto unknown iron chelator. PolyP might complex excess iron as it is being taken up from the extracellular space and/or released from iron-containing proteins during cisplatin treatment. We reasoned that if this model were to be correct, we should be able to specifically rescue the cisplatinsensitivity of the ppk deletion strain by reducing the intracellular iron load during cisplatin treatment. To test this idea, we devised two different strategies: (i) decreasing the extracellular iron concentration, which should reduce the amount of Fe uptake during cisplatin stress (36), or (ii) overexpressing the iron storage protein FtnA, which should compensate for the lack of polyP. For our first strategy, we grew wild-type and ppk deletion strains in liquid medium supplemented with either normal (ϩFe) or low (ϪFe) concentrations, exposed them to our previously established cisplatin treatment, and determined cell survival and mutagenesis rates. The result that we obtained was fully consistent with our hypothesis: lowering the iron concentration in the media dramatically increased the survival and reduced the mutagenesis rates of the ppk deletion strain to levels directly comparable to cisplatin-treated E. coli wild type (Fig. 3A  and B). In contrast, cisplatin treatment under low versus normal iron conditions did not yield any noticeable difference in the cisplatin resistance of wild-type E. coli. These results suggested that iron taken up from the media in response to cisplatin treatment was successfully complexed or otherwise neutralized by polyP. We obtained a very similar result upon overexpression of the plasmid-encoded iron storage protein FtnA, which significantly improved survival and reduced the mutagenesis rates in the cisplatin-treated ppk deletion strain but not in wild-type E. coli (Fig. 3C and D). These results demonstrated that expression of a protein-based iron chelator fully compensates for the lack of polyP and suggested that iron complexed in metalloproteins might contribute to the iron toxicity in the absence of polyP. Since previous reports documented that cisplatin covalently binds cysteine residues in proteins (9, 13, 37), including Exponentially growing E. coli wild-type (wt) and ppk deletion strains were exposed to 10 g/ml cisplatin in untreated (ϩFe) or Chelex treated (ϪFe) M9-G medium for 20 h to determine survival (A) or for 1 h to determine the rate of mutagenesis (B). (C and D) Exponentially growing E. coli wild-type and ppk deletion strains carrying either an empty plasmid or a plasmid overexpressing the iron storage protein FtnA were exposed to 10 g/ml cisplatin in M9-G medium for either 20 h to determine survival (C) or 1 h to determine the rate of mutagenesis (D). those involved in iron binding (38), we finally tested the idea that cisplatin targets iron-containing proteins in bacterial cell lysates. We therefore prepared wild-type or ⌬ppk crude extracts and measured the activity of aconitase, an enzyme, whose ironsulfur cluster is coordinated via three oxidation-sensitive cysteines. We found that increasing concentrations of cisplatin indeed increasingly reduced the activity of aconitase at similar levels in both wild-type and ⌬ppk extracts (Fig. S3). Together, these results suggested that endogenous polyP provides an effective nonproteinogenic mechanism to protect bacteria against cisplatin-mediated accumulation of free iron and Fe-mediated oxidative damage.
PolyP acts as a physiologically relevant iron storage molecule in vivo. Iron homeostasis is a tightly regulated mechanism, necessitated by the fact that unbound labile iron is highly toxic under aerobic growth conditions. This toxicity appears to be primarily caused by the ability of free iron to interact with peroxide (i.e., Fenton reaction), which leads to the production of highly reactive hydroxyl radicals (39). To further evaluate the idea that polyP serves as a general, hitherto unrecognized iron storage molecule in bacteria, we turned to a mutant strain of E. coli, which lacks the master iron repressor Fur. Deletion of Fur triggers an iron starvation response, which, similar to the situation we observed in cisplatin-treated wild-type E. coli, leads to gene expression changes that are geared toward replenishing the intracellular iron pool (40). As a cellular consequence, fur deletion strains suffer from an intracellular iron overload, which causes severe growth defects and significantly higher mutagenesis rates under aerobic but not under anaerobic growth conditions (28). To test whether polyP functions as an iron storage molecule in a ⌬fur deletion strain, we generated ⌬fur ⌬ppk double deletion strains and analyzed growth and mutagenesis rates under both aerobic and anaerobic growth conditions. As shown in Fig. 4A to C, additional deletion of ppk in the ⌬fur deletion strain significantly aggravated the growth defect and increased the mutagenesis rate specifically under aerobic growth conditions (Fig. 4A to C). These results were fully consistent with our prior observations and supported our conclusion that polyP protects bacteria generally against conditions of Fe overload. To finally test whether polyP also serves as an iron reservoir under nonstress conditions, we cultivated E. coli wild-type, ⌬ppk, and ⌬ppx strains in M9 minimal medium in the presence of increasing concentrations of the divalent iron chelator 2,2=-dipyridyl, using gluconate as sole carbon source. By offering solely gluconate, we ensure that bacterial growth depends on the activity of the iron-sulfur cluster protein gluconate dehydratase and hence on the availability of intracellular iron (41). Addition of 2,2=-dipyridyl to the growth media reduces the intracellular iron pools and, at sufficiently high levels, prevents bacterial growth as it depletes the Fe-S cluster in gluconate dehydratase. As shown in Fig. 4D (and Fig. S4A and B), whereas the ⌬ppx strain was slightly more resistant toward the presence of the chelator compared to wild-type E. coli, the ⌬ppk strain was significantly more sensitive. Less than 100 M 2,2=-dipyridyl in the growth medium was sufficient to decrease the relative growth rate of the ⌬ppk strain by 50% while more than 150 M chelator was necessary to trigger the same growth defect in wild-type E. coli (Fig. 4D and Fig. S4A and B). These results demonstrated that lack of endogenous polyP drastically increases the sensitivity of E. coli toward the presence of iron chelators in the media and supports the conclusion that polyP serves as a physiologically relevant iron storage molecule in E. coli.
PolyP protects against Fe-mediated DNA damage in vitro. To directly test whether polyP, through complexing iron, mitigates the Fenton reaction, we monitored the H 2 O 2 /FeSO 4 -mediated oxidation of 2,2=-azino-bis(3-ethylbenzothiazoline-6-sulfonic acid) (ABTS) (42) in the presence of increasing amounts of polyP (Fig. 5A). ABTS, once oxidized by H 2 O 2 /FeSO 4 -produced hydroxyl radicals, shows a strong absorbance signal at 414 nm. As shown in Fig. 5A, the presence of polyP prevented ABTS oxidation in a concentration-dependent manner, indicating that chelation of Fe 2ϩ by polyP inhibits the Fenton reaction. In fact, 20 M polyP 300 (in P i units) was sufficient to prevent oxidation of ABTS by a mixture of 5 M FeSO 4 and 20 M H 2 O 2 (Fig. 5A). To test whether the association of Fe 2ϩ with polyP also protects DNA against Fe-mediated oxidative damage, we incubated 10 M linearized DNA with a mixture of 50 M Fe 2ϩ /5 mM H 2 O 2 in the absence and presence of polyP 300 (Fig. 5B). Whereas in the absence of polyP, all of the DNA was oxidatively degraded within a 30-min incubation period, the presence of 5 mM polyP 300 (in P i units) almost completely prevented the degradation of DNA. The presence of inorganic phosphate (P i ) in the form of either sodium phosphate or potassium phosphate did not have any protective effect even when used at concentrations as high as 75 mM (Fig. 5B), confirming that the polyanionic nature of the polyP chain is necessary to sequester iron in a nonreactive state. In summary, these results demonstrate that polyP acts as a physiologically relevant iron storage molecule, capable of preventing the production of hydroxyl radical by the Fenton reaction.

DISCUSSION
Iron serves as an essential cofactor in many enzymes, which are involved in processes ranging from metabolism to DNA synthesis and cell division. Yet, free iron is highly toxic under oxygen-rich conditions as it readily undergoes the Fenton or iron-catalyzed Haber-Weiss reaction, thereby producing extremely reactive oxygen species, particularly hydroxyl radicals (39) (Fig. 6A). Because of this dichotomy in cellular need for and risk of free iron, aerobically growing organisms such as E. coli tightly regulate their iron homeostasis. Although the transcriptional regulation of iron homeostasis has been well studied over the years (43), little is known about potential first responders, such as physiological iron chelators, that could buffer the cellular effects of iron deregulation. Here, we demonstrate that cisplatin, an anticancer drug and broad- Experiments were performed in triplicates (for biological replicates, see Fig. S4A and B), and a representative result is shown. spectrum antimicrobial, triggers cytotoxicity in bacteria not only through its ability to cross-link DNA but also by causing cellular iron overload. By using a combination of genetic and biochemical tools, we discovered that bacteria defend themselves against this secondary insult through polyP, which effectively complexes the iron, prevents the Fenton reaction, and mitigates cisplatin toxicity (Fig. 6B).
It is well established that polyP interacts with metals such as Ca 2ϩ to form acidocalcisomes and protects organisms by complexing and sequestrating heavy metals (44)(45)(46). In contrast, however, very little is known about the interaction of polyP with  Our data suggest that cisplatin treatment causes iron accumulation that might contribute to a second mode of killing through Fenton-mediated oxidative stress and DNA damage. Iron stress likely results from a combination of events: the release of protein-bound iron and the inactivation of the master repressor of bacterial iron homeostasis Fur that triggers gene expression changes aimed to increase the cellular iron load. Subsequently, cellular free iron levels increase and can wreak havoc under aerobic growth conditions by producing highly reactive oxygen species, such as hydroxyl radicals. (B) Due to its polyanionic nature, endogenous polyP binds free iron and prevents iron from acting as a catalyst for the Fenton reaction, thereby drastically reducing the toxicity of cisplatin.
PolyP Functions as Fenton Reaction Inhibitor ® iron, the physiological relevance of such an interaction, or its potential role in the Fenton reaction (47,48). In fact, earlier in vitro studies suggested that polyP as well as shorter phosphate moieties, including pyro-, tri-, and tetrapolyphosphate, might actually stimulate the Fenton reaction (49,50). This result agreed with detailed follow-up studies, aimed to identify physiologically relevant Fenton inhibitors. This study revealed that nucleotide phosphates, such as ATP, either stimulate or prevent the Fenton reaction, depending on the number of iron coordination sites that were occupied by phosphates in the complex (48). The most effective Fenton reaction inhibitor in vitro turned out to be dimers of ATP-y-S and inositol-6-phosphate, which, through their six existing phosphate groups, block all iron coordination sites (48,51). This appears to prevent H 2 O 2 from reacting with iron and mitigates hydroxyl radical formation. We now propose that polyP, due to its polyanionic nature and structural flexibility, is able to also occupy all relevant coordination sites in iron, thereby interfering with peroxide binding and preventing the Fenton reaction (Fig. 6B). This feature would make polyP, to our knowledge, one of the few known physiologically relevant, nonproteinaceous, Fenton reaction inhibitors in living cells.
Our studies revealed noteworthy parallels between the protective function of polyP in bacteria exposed to cisplatin and in bacteria that lack the iron repressor Fur. Our transcriptional analysis of cisplatin-treated wild-type bacteria supported this conclusion by demonstrating that close to 80% of previously identified Fur regulons are upregulated in bacteria treated with cisplatin (22). At first glance, these results suggested that bacteria treated with cisplatin suffer from stress conditions that trigger intracellular iron depletion; hence, the upregulation of siderophores and transport mechanisms aimed to replenish the intracellular iron pools. However, analysis of the cisplatin response in polyP-depleted bacteria yielded quite the opposite result and, in fact, suggested that cisplatin triggers an accumulation of free iron, which becomes highly toxic and mutagenic unless chelated by polyP. Further support for this conclusion came from cisplatin treatment studies in Fe-depleted media, as well as complementation studies in which we overexpressed the E. coli Fe storage protein FtnA. In both scenarios, cisplatin treatment of the ppk deletion strain no longer increased the mutagenesis rate or affected survival beyond what we observed in cisplatin-treated wild-type E. coli. These results strongly argue that cisplatin treatment causes an increase in labile intracellular iron, which, unless chelated by polyP, significantly increases the toxicity of cisplatin. Since cisplatin is well known for its ability to interact with and bind to cysteine and methionine residues in proteins (11), and Fur contains a redox-sensitive cysteinecoordinating zinc site (52), we now speculate that Fur itself might become a target of cisplatin. Inactivation of Fur would misleadingly send an iron starvation signal to the cell, causing iron accumulation. Intriguingly, very recent studies showed similarly disruptive effects of cisplatin on the iron homeostasis in cancer cells (53). In contrast to bacteria, however, which ultimately suffer from iron overload, cisplatin-treated cancer cells experience true iron starvation. This iron starvation phenotype is triggered by the covalent modification of two cysteines in the iron regulatory protein 2 (IRP2), a central activator of the mammalian iron starvation response (53). Once inhibited, IRP2 is unable to downregulate the iron chelator ferritin, causing persistent iron depletion. Our realization that polyP serves as an iron chelator helps to explain our recent finding that endogenous levels of polyP positively correlate with apoptosis in cisplatin-treated cancer cells (16). We now reason that by chelating iron, polyP further potentiates the iron starvation phenotype in cancer cells, hence causing the observed increase in cytotoxicity (16,53).
The ability of polyP to complex iron might also contribute to its protective function under oxidative stress conditions, as well as other cellular insults that cause proteinbound iron to be released. Relying on nonproteinaceous compounds to chelate a toxic but essential metal might reflect a global strategy of the stress response. PolyP accumulation does not require time-consuming transcription or translation processes, and the molecule itself is stable toward almost all relevant damaging agents (18). Moreover, polyP ensures the proper activation of RpoS, the master regulator of the general stress response in E. coli leading to the expression of genes such as catalases to increase the overall resistance to oxidative stress (54). In addition, chelation of iron by polyP would allow iron to be rapidly reincorporated into proteins without the activation of complex and energy-consuming uptake systems once the stress is removed. It now remains to be tested, however, whether and how association of polyP with iron affects some of the other known functions of polyP, such as its ability to interact with and stabilize unfolding proteins, or its activity in blood coagulation and inflammation (55). This would add another exciting new layer of complexity to this structurally simple molecule.

MATERIALS AND METHODS
Bacterial strains and growth conditions. All strains, plasmids, and oligonucleotides used in this study are listed in Table S5 in the supplemental material. Gene deletions were generated by red-mediated site-specific recombination (56). All chromosomal mutations were confirmed by PCR. E. coli MG1655 was grown at 37°C in lysogenic broth (LB; Fisher) or MOPS minimal medium (Teknova) containing 0.2% (wt/vol) glucose and 1.32 mM K 2 HPO 4 (MOPS-G). To conduct cisplatin resistance tests on plates, M9 minimal medium (57) containing 0.2% glucose (M9-G) was supplemented with 1.5% agar. When indicated, 0.2% (wt/vol) gluconate was used instead of glucose (M9-gluconate). For iron-depleted conditions, M9-G medium was mixed and incubated with 2 g/liter Chelex100 (Bio-Rad) for 1 h at room temperature under constant agitation. The chelated solution was then sterile filtered. For pBAD expression induction, arabinose (0.2%, wt/vol) was added 15 min prior to the experiment in the case of liquid assays or added to the media in the case of plate assays. The following antibiotics were added when appropriate: chloramphenicol (30 g/ml), rifampin (200 g/ml), kanamycin (50 g/ml), or ampicillin (100 g/ml).
Growth under iron starvation. Iron starvation growth assays were performed as described in reference 58. Briefly, E. coli MG1655 wild-type and ppk and ppx deletion strains were grown overnight in M9-gluconate medium, diluted, and cultivated at 37°C until an OD 600 of 0.5 was reached. Cells were then diluted 1:100 into M9-gluconate medium supplemented with the indicated concentrations of 2,2=dipyridyl (stock dissolved in 100 mM dimethyl sulfoxide [DMSO]; Sigma-Aldrich). Each condition was performed in triplicate, and growth was monitored for 18 h. The OD 600 reached after 18 h was then normalized to the corresponding growth rate in M9-gluconate in the absence of chelator and plotted.
Cisplatin survival assay. E. coli MG1655 wild type and the isogenic mutant strains were grown at 37°C with shaking in MOPS-G medium to an OD 600 of ϳ0.4 to 0.8 and harvested by centrifugation. To analyze cisplatin sensitivity in liquid culture, the cells were resuspended in MOPS-G medium to an OD 600 of 0.4, supplemented with various concentrations of cisplatin (stock solution 0.9 mg/ml in sterile double-distilled water [ddH 2 O]; Sigma-Aldrich), and incubated at 37°C with shaking (200 rpm). At defined time points of incubation (1 h to 20 h), the cells were harvested by centrifugation, washed twice, 10-fold serially diluted, and plated onto LB agar. Survival was assessed after overnight incubation at 37°C. To determine the survival of bacteria when grown on cisplatin-containing plates, bacteria were cultivated in MOPS-G medium until an OD 600 of 0.5 was reached. Then, the bacteria were 10-fold serially diluted and plated onto M9-G plates containing various concentrations of cisplatin. Colonies were counted after 24 h of incubation at 37°C. To compare the cisplatin sensitivities of bacteria in iron-depleted liquid media, the bacteria were first grown in MOPS-G medium as described above. Once an OD 600 of 0.4 was reached, the bacteria were centrifuged, washed, and resuspended in either untreated M9-G (ϩFe) medium or Chelex-treated M9-G (ϪFe) medium in the absence or presence of 10 g/ml of cisplatin. At defined time points of incubation (1 h to 20 h), the cells were harvested by centrifugation, washed twice, 10-fold serially diluted, and plated onto LB agar.
Mutagenesis assay. Mutagenesis rates were measured as described previously (59). Briefly, cells were grown overnight in MOPS-G medium, diluted 1:100 into 30 ml fresh MOPS-G medium, and cultivated at 37°C until an OD 600 of 0.5 was reached. The bacterial suspension was either left untreated or supplemented with 10 g/ml cisplatin and further incubated for 1 h. After the incubation, untreated and treated bacteria were washed twice in MOPS-G medium and resuspended in 5 ml of LB medium. The bacteria were then incubated overnight at 37°C with shaking. Serial dilutions were made and plated onto both LB agar and LB-rifampin agar plates. After 24 h of incubation at 37°C, the CFU were scored. The mutation frequency was calculated by dividing the number of rifampin-positive colonies by the total number of colonies.
RNA-seq analysis. Four biological replicates of wild-type E. coli MG1655 and ⌬ppk mutant were cultivated in MOPS-G medium at 37°C until an OD 600 of 0.4 to 0.5 was reached. Cells (1 ml) were harvested either before or 15 min after the treatment with 20 g/ml cisplatin in 1 ml of ice-cold methanol (Ϫ80°C) to stop transcription. After centrifugation and removal of the supernatant, total RNA was prepared using the Ambion RiboPure-Bacteria kit (Thermo Fisher Scientific) according to the manufacturer's instructions. The samples were DNase I treated, followed by depletion of rRNA using the Illumina Ribo Zero kit (Illumina) for Gram-negative bacteria. Fifty-base single-end sequencing was performed on an Illumina HiSeq 4000 using the University of Michigan DNA Sequencing Core. Sequence reads from the RNA-seq were mapped onto the reference genome (NC_000913). Genes with a log 2 fold change of Ͼ0.5 and a false-discovery rate (FDR) value of Ͻ0.01 were considered differentially expressed genes (DEGs).
COG enrichment and analysis. Differentially regulated genes upon cisplatin treatment and/or between strains were categorized according to their annotated COG categories (30). Functional enrich-ment of COG categories was determined by performing a modified one-tailed Fisher exact test (EASE score from DAVID), with a P value of Ͻ0.05 considered significant. Comparison to iron-regulated genes (22) seen in the Venn diagram in Fig. 2C was performed using a Fisher exact test. The heatmap was produced using open-source statistical software R (https://www.r-project.org/) with log 2 fold change data. Figures 2 and S2 were also produced in R.
Enzymatic assays. The aconitase activity assay was performed according to the manufacturer's protocol (MAK051; Sigma-Aldrich). Briefly, bacterial cell pellets were resuspended in 1 ml of aconitase lysis buffer (0.1 mM Tris-HCl, pH 8.0, 0.1 M KCl, 1 mM phenylmethylsulfonyl fluoride [PMSF], and 0.6 g/l lysozyme) and lysed by 5 to 6 rounds of freeze-and-thaw cycles. The lysates were centrifuged at 14,000 rpm for 10 min at 4°C. The total protein concentration was determined using the Bradford assay. The aconitase assay was performed by mixing 100 g of cell lysate with 200 l of 1ϫ aconitase assay buffer (0.6 mM MnCl 2 , 25 mM sodium citrate, 0.25 mm ADP, 50 mM Tris-HCl, pH 7.6). The absorbance was recorded at 340 nm, and the specific aconitase activity was calculated per milligram of total proteins.
In vitro Fenton reaction and DNA damage assay. The ABTS assay was performed as previously described in reference 42. In brief, 250 M ABTS was mixed with 5 M FeSO 4 and 200 M H 2 O 2 in the absence or presence of increasing concentrations of polyP 300 (see figure legends for details). The assay was performed in 10 mM acetate buffer, pH 3.6, to control acidity. After 30 min of incubation at 37°C, the mixture was read at an absorbance of 414 nm. The DNA damage assay was performed by mixing 50 M FeSO 4 , 5 mM H 2 O 2 , and 10 M linearized DNA from the plasmid pBAD18 in water with the indicated concentration of either Na 2 HPO 4 , KH 2 PO 4 , or polyP 300 (concentration given in P i units). The reaction was initiated by the addition of H 2 O 2 , and the reaction mixture was incubated for 30 min at 37°C. Then, samples were loaded onto an agarose gel. Staining with ethidium bromide was used to visualize the DNA bands.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. TEXT S1, DOCX file, 0.02 MB.